A Method for Medical Image Visualization: Integrating Eye-Tracking for Pixel Error Reduction and Calibration

Clinical procedures depend heavily on visual data, especially in diagnostic radiology. Using eye-tracking technology, this study examines the association between visual search features and diagnostic performance in capsule endoscopy. Conventional visual search models may not be appropriate for cross-sectional stack imaging used in computed tomography (CT) and magnetic resonance imaging (MRI) due to their dynamic nature. These models were created for static 2D pictures. By examining the visual search patterns of a specialist looking at capsule endoscopy pictures, our study seeks to close this gap. To show these pictures, we created a framework that included an eye tracker and gathered a sizable dataset of 536 photos. The eye-tracking system demonstrated high accuracy in capturing gaze data. Analysis of expert visual search patterns revealed predominantly circular search trajectories with distinct optical flow characteristics. Variations in fixation duration and location were observed. Notably, visual search strategies resembled previously documented drilling and scanning patterns in certain image types. These findings suggest the potential of utilizing eye-tracking data to enhance diagnostic accuracy and inform radiology education.